The Digester

AI could add objectivity to depression diagnosis, but risks remain

Mar 20th 2026

With depression rates rising and few treatment advances since the 1980s, researchers hope AI can identify objective biomarkers from speech and facial cues to improve diagnosis and treatment matching, but bias, errors and recent failures on women’s health queries mean robust protections are essential.

  • Depression is still diagnosed with vague symptom checklists rather than clear biomarkers.
  • Researchers are training AI to detect physical signals like facial expressions and voice cadence to create objective biomarkers.
  • A recent review found exercise can be as effective as antidepressants or cognitive behavioral therapy for treating depression.
  • AI tools carry known risks including training data bias and hallucinations, and a study found top models gave inadequate advice for 60 percent of women's health queries.
  • Strong safeguards and validation are required before AI is used to guide diagnosis or match patients to treatments

Articles